Toward risk adjustment in mental health in Israel: calculation of risk adjustment rates from large outpatient and inpatient databases

Mental health services in Israel are included in the third addendum to the National Health Insurance Law (1994) [1], along with other services that until the law came into effect were provided by the state. Accordingly, most ambulatory and hospitalization services were provided in the past by government operated mental health clinics and hospitals, rather than by the Health Plans. This dichotomy of separation of psychiatric and other mental health services from the rest of health care is undesirable in the opinion of the majority of those concerned, primarily with regarding patients who suffer from psychiatric disorders. The inclusion of mental health in the basic health package provided by the Health Plans is essential for the following reasons:

Since the enactment of the National Health Insurance Law, several failed attempts were made to include mental health services as part of the health package of services that the Health Plans are obliged to provide (“the mental health insurance reform”). In 2003, a decision to do so was made by the government, following which negotiations were held between the Ministries of Health and Finance and the rest of the parties involved, and in 2008–2012 an effort was made to implement the change in the framework of the legislation [5]. These efforts encountered many difficulties, among others, the objection of professionals who feared that the quality of service provided to patients would be impaired, and who suggested that government owned clinics should remain in their current format (the “benevolent reform”) [6]. In the end, a regulation was issued by the Minister of Health to include mental health services in the health services package as of 1/7/15 [7].

Towards the inclusion of mental health in the health services package, the Ministry of Health and the Health Plans began the process of evaluating and establishing new services [8]. For the purpose of expansion, the services package budget was increased by NIS 1.261 billion for hospitalizations and NIS 0.776 billion for ambulatory services.

Along with the clinical and therapeutic issues arising from this change, the issue of the allocation of the (additional) budget to the Health Plans should be considered. A new risk-adjustment (capitation) formula which covers (both) physical and mental health care is needed. The inter-ministerial team which was established for this issue [9] recommended allocating the mental health hospitalization funds according to the historical and expected share of the Health Plans in national mental health hospitalization expenditures. These calculations take into account the current process of decrease in the relative share of the largest Health Plan, “Clalit”, and increase in the share of the smaller Plans over the years (Table 1). The budget for ambulatory services is added to the existing (physical health care only) budget, which is allocated among the Plans by adding to the current risk-adjustment formula a new expenditure category – ambulatory mental health care - based on age only. This was done taking into consideration the goal of increasing usage of ambulatory services by 2% in persons below the age of 24, and 4% in persons above 25. Table 2 shows the changes in the pre-reform capitation rates, i.e., the mean cost of care in the risk-adjustment age-gender group relative to the overall mean cost of care in the population, implied by adding ambulatory mental health care as a separate "expenditure category. Basically, the change transfers funds that go to the Health Plans for insurees in advanced age to insurees in middle age.

The inter-ministerial team’s risk-adjustment scheme has several shortcomings, most of them originated from lack of data. First, while payment for ambulatory mental health care is integrated into the payment for care provided by the Health Plans, it is done using the methodology of “expenditure heads”, a methodology which is not the common one used elsewhere and which is subject to major criticism. The major drawback is the assumption of equal unit-costs across health plans [10]. Second, the only risk adjuster used for ambulatory mental health care is age. Diagnosis, marital status and sex, for example, are not included although they exercise significant effects on mental health services use. Third, the payment for inpatient mental health care is determined by the historical and predicted shares of the Plans in mental health hospitalization. This procedure presents no incentives towards any social goal. Furthermore, it is not clear why the predicted share – even if accurate – indicates the (predicted) cost of care, since the Health Plans can substitute outpatient care for inpatient care and physical health care for mental care. The recommended risk-adjustment scheme is thus clearly insufficient in preventing risk selection, skimping on quality and quantity of care and distortion of the competition among the Health Plans [11].

The international experience with risk adjustment in mental health services shows a weak-moderate prediction of psychiatric services consumption based on different a variety of demographic and clinical variables. The RAND study examined a subset of insured patients in California with high expenditures on mental health during the years 1991–1994 [12]. Gender, age, ethnic origin, marital status, entitlement to social services, psychiatric diagnoses, a functional index and past expenditure, together predicted up to 16% of the variance of expenditure on mental health. Another study from the Veterans Administrations in the US [13] found that gender, age, medical and psychiatric diagnoses, and alcohol and drug abuse, predicted up to 20% of the variance of the expenditure on mental health and 32% in general health expenses. In the Netherlands, a separate risk-adjustment scheme was in effect until last year, when a reform combined physical and mental health care.

The risk-adjustment rates signify the mean cost of care in the risk-adjustment group (defined by the choice of the risk-adjusters) relative to the overall mean cost of care in the population. The derivation of the risk adjustment rates in this study consists of two steps: first, based on utilization data and the selected risk adjusters, the rates are calculated for the users, namely, those who have a non-zero cost of care. Second, in each group, the rates are recalculated taking into account the number of non-users, namely those with zero cost. Because of limited data on the prevalence of the use of mental health care in some groups defined by the risk adjusters above, the rates were calculated for a restricted model only.

In order to calculate the rates among the users, we used results from our previous study [14]. We collected data on all patients hospitalized in psychiatric wards in Israel in the years 2012–2013 (n = 27,446) available through the Israel Psychiatric Case Registry [15]. Days of stay were calculated only for the 2 years studied even if hospitalization started before or ended after this period. We computed the total expenditure on hospitalization for each patient.

Data on all patients receiving treatment in one psychiatric outpatient clinic in Kiryat HaYovel, Jerusalem (n = 6115) were also available. Although we can not argue that this clinic is an accurate representative of all mental health clinics in Israel, it does serve a heterogenous population of Jews and non-Jews, high and low socioeconomic groups and urban and rural populations residing in West Jerusalem and the vicinity. For these patients we were able to calculate expenditure on ambulatory mental health care. This was done by multiplying numbers of ambulatory visits by the respective Ministry of Health tariff. Regression models were used to predict expected expenditures. Three models were estimated: the “demographic model” (including age and gender as risk adjusters), the “socio-economic model” (where socio-economic variables were added), and a “clinical model”, where diagnoses were added [14].

Because of the limited data on use in the groups defined by the models above, for the calculation of the rates for ambulatory care, we used the “demographic model”. For inpatient care, a “restricted demographic-socioeconomic-clinical” model was constructed. This model includes gender and age groups as well as the two strongest predictors we identified in our analysis: whether the person is single or not and has a schizophrenia spectrum diagnosis (SSD) or not. SSD was the most prevalent diagnostic group in inpatients in our study (51% compared to 1.6–15% for other diagnostic groups). Also, much less is known about the prevalence of the other diagnostic groups in different age and gender groups and their relationship to being single, data that were needed for calculations of weights.

Data on the number of hospitalized patients in each age and gender group in the study period were taken from the Israel Psychiatric Case Registry. Data on the number of patients in public mental health clinics in Israel during the study period in each age – gender group were obtained from the Ministry of Health Statistical Report for 2012 [16]. Data on the number of individuals residing in Israel in the different age and gender groups were taken from the Statistical Report of the Central Bureau of Statistics for 2012 [17].

For SSD it was possible to estimate the size of the relevant subgroups in the general population, based on data from the literature: The prevalence of schizophrenia in the general population is 1%, the average age of onset is 10–25 for men and 25–35 for women [18]. The prevalence of marriage among schizophrenia patients is lower than the average in the population. Data on the prevalence of marriage among schizophrenia patients were found for various developing countries in the world. No data were found for developed countries. A proportion for 20% of married individuals among individuals diagnosed with schizophrenia was chosen according to data from Sao Paulo, Brazil, the population with the closest socioeconomic status to Israel among those reported [19]. These estimates of schizophrenia and single status in schizophrenia were used to assess the size of subgroups in the population for different age groups and genders. The number of hospitalized patients in each subgroup was taken from the Psychiatric Case Registry data for the study period.

The ambulatory and inpatient capitation rates were weighted by the share of each service in total mean mental health care cost in the population as found in the study (NIS 12 and NIS 166 respectively) and by their shares in the budget as determined by the inter-ministerial team (1.8 and 2.8% respectively) to create capitation rates for (total) mental health care. These latter rates were further combined with the pre-reform (physical health care) capitation rates using their shares in mean cost, NIS 178 for mental health (this study) and NIS 4554 for physical health (for the year 2014), and in the budget (4.6 and 95.4% respectively) to produce overall post-reform risk adjustment rates. Finally, the shares of the Health Plans in the post-reform budget – determined by their shares in total “standardized insurees” according to the age-gender composition of their insured population (data from the National Insurance Institute for November 2015) - were calculated for both the inter-ministerial team’s recommendations and our overall capitation rates.

Based on previously identified predictors of mental health expenditure we were able to calculate rates for the capitation formula by age and gender in ambulatory care. For inpatient expenses, the rates were calculated adjusting for gender, age, being single, and a diagnosis of schizophrenia spectrum disorders. The range of values ​​of the calculated rates illustrates how different the risk may be between different groups of insured individuals.

The rates we calculated are different from what was proposed by the inter-ministerial team that dealt with this subject. The team did not suggest rates for the hospitalization expenses, and as can be seen from our results these vary widely between different age and gender groups, and much more when a schizophrenia spectrum diagnosis and singlehood are added.

Failing to account for that heterogeneity might lead to risk selection, skimping on quality and quantity of care and distorts the competition among the Health Plans.

Two limitations of the study should be mentioned: The calculation of the rates for the entire population (including non-users) is based on “imported” parameters, which might not reflect the Israeli reality. Second, our ambulatory scale is based on one – not necessarily representative – clinic.

The risk adjustment rates we calculated based on individual data are substantially different from the ones suggested by the inter-ministerial team that dealt with allocation funds for mental health in Israel. The inter-ministerial team chose to allocate the inpatient mental health funds according the share of the Health Plans in total mental hospitalization historical cost, which implicitly assumes that all mental health inpatients and days are similar. Since this assumption is clearly violated in practice, it creates incentives to risk selection and distortions.

While this study suffers from several limitations, it sheds some insights into the mental health risk adjustment scheme. In particular, it shows the need to consider diagnosis (especially SSD) as risk adjustment. It seems that the care provided to SSD patients should be compensated separately (e.g. as a “severe condition”), which was not done yet. At present, after four years of practice under the reformed system, Health Plans individual data should be used to calculate a modified post-reform risk adjustment rates.